109
Initial Results of Calibrating the Baron LIS-NOAH V2 Fully Distributed Hydrological Modeling System for the DMIP Elk River Basin

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 19 January 2010
John McHenry, Baron Advanced Meteorological Systems, Raleigh, NC; and D. J. Gochis, D. N. Yates, and C. J. Coats Jr.

Handout (527.6 kB)

The Baron LIS-NOAH Version 2 (LN2) is a fully distributed/routed hydrological modeling and forecasting system containing a number of state-of-science sub-models: (1) the NOAH-Unified Land-Surface Model; (2) a 2D Diffusive-wave overland flow model; (3) a quasi-3D sub-surface saturated zone lateral flow model; (4) a sub-surface bucket-style baseflow model; (5) a diffusive wave channel routing model; and (6) a level-pool-based lake and reservoir model. The modeling system is being deployed operationally in Romania under the DESWAT program for 11 major basins there, with further deployments for other locations currently in the planning stages. As with most hydrological models, calibration is an important step in obtaining acceptable performance--even more so when operational hydro-meteorologists are relying upon forecast outputs.

Because the OHD-based DMIP and DMIP-II programs offer very high quality long-term hydrometeorological datasets, the development of a calibration protocol for LN2 is relying in part on this resource. In particular, the Elk River basin near the Missouri-Oklahoma-Arkansas border is being used to benchmark an initial calibration effort. The effort is targeting a systematic approach in which (1) land-surface, overland flow, and subsurface flow algorithms are first calibrated for long-term observed stream inflow; (2) the baseflow/bucket component is calibrated using long-term observed baseflow; and finally (3) the channel routing is calibrated to improve hydrograph timing and peak flows. In each case, the non-linear PEST package is providing the optimization kernel. Spin-up, calibration, and validation time-periods are being chosen in order to isolate the observed data used for validation from that used for calibration.

Because such calibrations almost always require long-term datasets (order of a decade), the time scale of relatively rapid "greenhouse-gas-based" climate-change may potentially approach that same length, where hydrological forcing inputs are concerned (i.e. precipitation and temperature). This means that the need to perform repeated calibrations of most operational hydrological models can only increase if climate change intensifies. Hence, new strategies may have to be adopted whereby a nearly continual process of long-term calibration of operational models is under way in something of a "revolving door" fashion at operational centers.

This talk will report on the initial calibration results and also address in general terms the kind of strategies that may be needed for "quasi-continual recalibration" of a model such as LN2 under relatively rapid climate change scenarios.